Tech founder and CCO says AI can run the funnel, but humans still close the deal

Aryan Mahajan is the Chief Commercialization Officer at Gerra and founder of Zoro, where he builds AI operating layers for revenue organizations.

We sat down with Aryan to learn how he's using AI to shake things up in revenue organizations. Here's what he told us.

Being his own case study

I'm Aryan Mahajan. I'm the Chief Commercialization Officer at Gerra. I'm also the founder of Zoro, where I build the private AI operating layer that runs revenue organizations.

I didn't arrive here from the top of a sales org. I started with $500 and no safety net, forced to make a tiny amount of money behave like a large team. That constraint became the whole philosophy: when you can't hire, you stop asking "Who do I add?" and start asking "What can one operator plus a fleet of AI agents run?" The answer became a company that went from $0 to $90K in MRR in 14 months and now generates $150K a month at margins above 95%. It operates with more than 30 distinct AI agents, a team you can count on one hand, and a content engine that has generated more than 50 million views and 50,000 leads.

I care about AI in revenue because I'm not theorizing about it. I'm the case study. The revenue organization I lead is proof that the operating layer I build for others works.

Overhauling the top of the revenue funnel

I replaced the entire top of my revenue funnel — prospecting, research, qualification, and first touch — with an AI system that runs the commercial sequence end-to-end.

Before, human hours were the expensive front of revenue. A sales rep spends roughly 40 minutes researching a single prospect. A team of four or five costs $100-200K a year. Most of that work involves reading, guessing fit, and writing a first message. This work is judgment-heavy but slow, inconsistent, and capped by the number of humans you can afford.

I built a pipeline that sources prospects, enriches each one, researches the company and the specific buyer, scores fit against the ideal-customer profile, kills the ones that don't belong, writes the outreach in multiple languages, and then handles the replies.

Research that took 40 minutes per lead now costs about two cents per lead and runs 24/7. The economics inverted: a one-time system build replaces a $100K-200K per year in staffing costs, and the client owns the system outright. Because it logs every reply and outcome, the revenue motion finally has memory: what worked, what didn't, and why, instead of that context living in five different reps' heads.

The clearest proof: on one campaign, it researched 214 prospects, killed the 184 that didn't fit, and got a $70 billion hedge fund to reply while I was sitting on my couch. That's not "AI writing cold emails," which was interesting in 2022. That's AI running the entire commercial sequence. That's the shift. But most revenue leaders are still bolting a chatbot onto a workflow that shouldn't exist and calling it transformation.

Why taste and judgment must remain human

I let AI power the entire research-and-execution layer: prospecting and enrichment, ICP scoring and lead qualification, first-draft outreach and content, reply classification, pipeline and revenue monitoring — and even system repair. I run a self-healing loop that watches my live client systems, catches a failure, and fixes it autonomously; I've clocked a bug going from reported to resolved in four and a half minutes. AI handles anything high-volume or pattern-based, particularly work that improves as the system receives more context.

Tasks that remain explicitly human include pricing, positioning, final commercial decisions, and judgments about taste. I price based on the leverage a system creates, not its running cost. I do not hand that judgment — what an outcome is worth to a specific buyer — to a model. The same applies to the final words on a deal and voice. AI made output free, making taste the expensive thing. Everyone now has the same models, so the market quietly exposes who actually has judgment. I refuse to automate that.

Nor should AI close deals. Real people build trust in a room or a DM, and that trust has led to some of my biggest contracts. The last mile of revenue is about relationships and judgment, and AI has barely moved it.

My dividing line is simple: AI decides nothing that carries my reputation or commits me to a buyer. AI informs and drafts everything, but a human approves anything that goes out under my name or moves money. My system researches, scores, and writes the message for a prospect, but a person approves the send. It drafts a client reply, but I tap to send.

Why AI can't create demand on its own

AI doesn't manufacture authority or demand on its own. It can produce infinite content, but volume is not distribution, and it isn't trust.

I've generated 50M+ views, and the views are not where the money is — the money is in the handful of relationships those views made possible. Teams that expect AI to "generate pipeline" by producing more output usually just get more noise.

Demand comes from a point of view and real proof — human inputs the model amplifies but doesn't originate.

Why revenue leaders must be wary of overreaching automations

I see two big downsides to AI in revenue.

First, it can be confidently wrong. Early versions of the outreach system targeted hallucinated company domains until I required each one to be verified before use. The fix was not a smarter model, but a verification gate.

Second, automation that overreaches is dangerous. An early autonomous payment run double-processed a vendor payment based on a stale approval. Now, anything that moves money or goes to a client requires a hard approval gate and an undo window — the system drafts and prepares, but a human commits.

And I think it's important to note that the bottleneck has simply moved; it did not disappear. When output becomes free, taste, judgment, and guardrails — the human review layer — become the constraint, not production. Teams that bolt AI on without rebuilding that layer just generate more wrong things quickly.

Aryan Shares

I think it’s important to note that the bottleneck has simply moved; it did not disappear. When output becomes free, taste, judgment, and guardrails — the human review layer — become the constraint, not production.

How revenue teams are changing

AI has made the team smaller, more generalist, and more judgment-heavy. I do not need a human seat for every production step. I need a small number of operators who understand the customer, the economics, and the system well enough to direct machines and own the exceptions.

I have built more than 30 distinct agents for research, drafting, monitoring, follow-up, delivery, and repair. Human effort is concentrated where the machine is weak: positioning, pricing, relationships, taste, accountability, and deciding what should happen when the standard playbook breaks.

The talent premium has shifted toward operator-builders — people who can sit in a commercial conversation, understand the business consequence, and then shape the workflow or work with engineers to change it. Pure coordinators have less leverage because AI can move information and produce the first pass. Pure specialists still matter, but their judgment must influence the broader organization rather than remaining limited to their own output.

The goal is not to remove humans from revenue. It is to stop spending your best humans on reading, copying, chasing, and updating, then put them back on trust and decisions.

Why CROs must rebuild their processes

I have three pieces of advice:

1. Stop bolting AI onto existing processes; rebuild them instead. Most "AI transformation" in revenue involves gluing a chatbot to a workflow that shouldn't exist—and if the process is wrong, AI only makes the wrong thing faster. Delete the step, then automate what's left. Organizations that treat revenue as an infrastructure problem — wiring authority, acquisition, memory, follow-up, and attribution into one system — compound their advantage while others still write prompts.
2. Own the layer, don't rent it. Software has become commoditized; a team with full operating context and real distribution consistently beats out-of-the-box SaaS. The winning move isn't buying ten more tools; it's building the operating system beneath your revenue organization so intelligence compounds, and you own it. Buy tools for commoditized functions; build the part that is your actual edge.
3. Become obsessive about the human layer, because that's where your advantage has shifted. AI made output free, making taste, judgment, and trust scarce and expensive. Direct AI at volume — research, drafting, monitoring, follow-up — and concentrate your best people on the last mile it cannot touch: pricing, positioning, relationships, and closing. And implement a strict approval gate on anything that goes to a customer or moves money. The goal isn't fewer humans making more noise; it's scaling your judgment.

The CROs who win this moment won't be those who adopt the most AI. They'll be those who rebuild their revenue organizations around it while it's still uncomfortable.

Follow along

You can follow Aryan Mahajan's work on LinkedIn, Instagram, and X. Or book a call with him.

More expert interviews to come on The CRO Club!